【Objective】Fully exploiting the spatial correlation of passenger flow between related stations in the subway network has a positive effect on the improvement of subway passenger flow prediction accuracy. Capturing and quantifying spatial patterns in passenger flow data is difficult due to the difficulty of learning and transferring spatial correlations between metro stations. 【Method】An improved graph-convolution gated recurrent neural network metro passenger flow prediction model is proposed to enhance the model"s ability to handle different data types by integrating multivariate spatio-temporal data. A spider wasp optimisation algorithm based on Tent chaotic mapping and Levi"s flight perturbation strategy is used to dynamically adjust the model structural parameters in order to optimise the hidden layer structure of the gated recurrent neural network.【Result】Prediction results on weekdays and weekends show that 20 iterations lead to optimal results, with higher prediction accuracy with larger training samples. 【Conclusion】Dynamic optimisation of the hidden structure of gated recurrent networks can lead to better convergence of the prediction model and higher prediction accuracy.